IndianDeepGuard: Enhancing Deepfake Detection through Cultural and Multi-Modal Context

DOI

10.2991/978-94-6239-791-0_8How to use a DOI?

Keywords

Deepfake detection; multi-modal learning; domain adaptation; cultural context analysis; CLAHE; DANN; Indian multimedia

Abstract

Deepfake technology has advanced rapidly, threatening information integrity across diverse demographics. This paper proposes IndianDeepGuard, a novel multi-modal framework designed to address generalization gaps when Western-based deepfake detectors are applied in the Indian context. The framework compares GenConViT, EfficientNetB4+Bi-LSTM, ResNeXt-101+Bi-LSTM for spatial-temporal feature learning, and incorporates cultural modifications such as Contrast Limited Adaptive Histogram Equalization for illumination normalization, a Cultural Context Extractor for analyzing cultural attire, and Multilingual Acoustic Anchoring (Wav2Vec2-XLS-R) for phoneme-viseme learning. Domain-Adversarial Neural Networks enable device-invariant feature learning. ResNeXt-101 achieves peak accuracy of 96.10% with a Macro-F1 score of 94.22%, outperforming the baseline (58.00%) by a substantial margin. EfficientNet-B4 achieves 95.67% accuracy, while GenConViT attains 93.00%. Ablation analysis confirm that Indian-specific modules consistently improve performance across all models.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Kiah Monpara
AU  - Jeesha Savla
AU  - Mishree Sanura
AU  - Manan Shah
AU  - Prachi Tawde
AU  - Neha Agarwal
PY  - 2026
DA  - 2026/10/09
TI  - IndianDeepGuard: Enhancing Deepfake Detection through Cultural and Multi-Modal Context
BT  - Proceedings of the International Conference on Artificial Intelligence and Machine Learning for Agro-Ecosystems (ICAIML-AE 2026)
PB  - Atlantis Press
SP  - 75
EP  - 85
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-791-0_8
DO  - 10.2991/978-94-6239-791-0_8
ID  - Monpara2026
ER  -